Multi-Task Deep Residual Echo Suppression with Echo-aware Loss

TL;DR

This paper presents a multi-task deep residual echo suppression method using an echo-aware loss function, combining linear AEC and neural post-filter with TDE, achieving top rankings in the ICASSP 2022 AEC Challenge.

Contribution

It introduces a hybrid linear and neural echo suppression approach with a multi-task learning framework and an echo-aware loss, improving perceptual quality and echo suppression performance.

Findings

TDE module improves perceptual quality.

Using linear echo as input yields better results.

Adaptive filter convergence enhances performance.

Abstract

This paper introduces the NWPU Team's entry to the ICASSP 2022 AEC Challenge. We take a hybrid approach that cascades a linear AEC with a neural post-filter. The former is used to deal with the linear echo components while the latter suppresses the residual non-linear echo components. We use gated convolutional F-T-LSTM neural network (GFTNN) as the backbone and shape the post-filter by a multi-task learning (MTL) framework, where a voice activity detection (VAD) module is adopted as an auxiliary task along with echo suppression, with the aim to avoid over suppression that may cause speech distortion. Moreover, we adopt an echo-aware loss function, where the mean square error (MSE) loss can be optimized particularly for every time-frequency bin (TF-bin) according to the signal-to-echo ratio (SER), leading to further suppression on the echo. Extensive ablation study shows that the time delay estimation (TDE) module in neural post-filter leads to better perceptual quality, and an adaptive filter with better convergence will bring consistent performance gain for the post-filter. Besides, we find that using the linear echo as the input of our neural post-filter is a better choice than using the reference signal directly. In the ICASSP 2022 AEC-Challenge, our approach has ranked the 1st place on word accuracy (WAcc) (0.817) and the 3rd place on both mean opinion score (MOS) (4.502) and the final score (0.864).